基于优化SPOT和D-S证据理论的测试性验证方案

针对现有基于序贯验后加权检验的测试性验证方案对测试性设计指标之间的模糊参数空间考虑不足，以及未能充分运用测试性多源先验信息的问题，提出一种优化序贯验后加权检验和D-S证据理论相结合的测试性验证方案。首先，考虑测试性设计指标之间的模糊参数空间，构建三参数空间复杂假设，并基于Bayes理论研究序贯决策规则，同时确定决策因子以及决策阈值；其次，以测试性指标构成的参数空间为辨识框架，分别构造基于专家信息以及测试性试验数据等先验信息的基本信任分配函数，建立融合多源先验信息的优化序贯验证方案；最后，结合实例进行研究，并与经典验证方案、传统Bayes验证方案、序贯概率比检验方案以及序贯验后加权检验方案进行了对比分析。结果表明，该方案由于考虑了模糊参数空间以及充分融合了多源先验信息，有效解决了模糊参数空间的处理问题，同时所确定的平均故障样本量在决策支持的参数空间均优于其他方法。

Testability verification scheme based on optimized SPOT and D-S evidence theory

The existing testability verification scheme, which is based on the sequential posterior odds test, has insufficient consideration of the fuzzy parameter space between the testability design indicators, and it fails to make full use of the testability multi-source prior information. So a testability verification scheme based on the optimized sequential posterior odds test and D-S evidence theory is proposed. Firstly, when the fuzzy parameter space between testability design indicators is taken into consideration, a three-parameter space complex hypothesis is constructed. At the same time, the decision rules of the hypothesis test, decision factors and thresholds are proposed based on Bayes theory. Secondly, the parameter space composed of testability indicators is used as the identification framework, and the basic trust distribution functions based on multi-source prior information, such as expert information and testability data are constructed. Meanwhile, an optimized sequential verification scheme that integrates multi-source prior information is established. Finally, the research is carried out with examples, and the empirical results are compared with the classic verification scheme, the traditional Bayes verification scheme, the sequential probably rate test scheme, and the sequential posterior odds test scheme. The results show that the proposed scheme considers the fuzzy parameter space and fully integrates multi-source prior information, so it can effectively solve the problem of processing fuzzy parameter space, and the average fault sample size determined is better than other schemes in the parameter space of decision support.